Skip to content

Remove the automatic postprocess failure-policy machine - #11

Merged
tae2089 merged 1 commit into
refactor/remove-quality-analyzersfrom
refactor/remove-postprocess-policy
Jul 11, 2026
Merged

Remove the automatic postprocess failure-policy machine#11
tae2089 merged 1 commit into
refactor/remove-quality-analyzersfrom
refactor/remove-postprocess-policy

Conversation

@tae2089

@tae2089 tae2089 commented Jul 11, 2026

Copy link
Copy Markdown
Owner

What

Removes the automatic postprocess failure-policy layer. Postprocess steps (stored-flow + search-index rebuild) still run on every build/update — only the escalation machinery on top of them is deleted: auto fail_closed after repeated failures, per-namespace run-log history, and the two inspection tools. Step failures stay visible (warn log + failed_steps in the response with status: "degraded"), so nothing is silently swallowed.

Stacked on the analyzer-removal PR — review that first.

Changes

  • Delete internal/postprocess/policy and its two GORM models.
  • Migration 000007 drops ccg_postprocess_policy_state + ccg_postprocess_run_logs (sqlite + postgres); RequiredSchemaVersion 6 → 7.
  • Remove get_postprocess_policy and reset_postprocess_policy MCP tools (26 → 24).
  • Drop the postprocess_policy request arg and the fail_closed path; keep inline degraded behavior in build_or_update_graph / run_postprocess.
  • Remove the PostprocessPolicy Deps field, runtime wiring, HTTP /status postprocess summary, and the ccg status postprocess section (--errors / --recent flags).
  • Delete guide/postprocess-failure-policy.md (EN + ko) and sweep doc references.

Verification

  • go build + go vet clean; go test -tags fts5 ./... green (one pre-existing timing-flaky webhook test, TestStatusHandler_RecentRepos_FailureHasErrorFields, passes in isolation and per-package — unrelated to this change).
  • New test TestRunMigrations_SqliteDropsPolicyTablesAtHead asserts both tables are gone at head schema.
  • Real binary: ccg migrate (schema v7) → ccg build → ok; no postprocess tables created.
  • Real MCP server (streamable-http): tools/list returns 24 tools; get_postprocess_policy / reset_postprocess_policy absent; startup logs prompts=4.

Stacked on #10.

🤖 Generated with Claude Code

Postprocess steps (stored-flow + search-index rebuild) run on every
build/update; the escalation layer on top of them — auto fail_closed after
repeated failures, per-namespace run-log history, and two inspection tools —
was operational overhead the project's purpose does not need. Step failures
are already visible: they warn to the log and list in the `failed_steps`
response field with `status: "degraded"`.

- delete internal/postprocess/policy and its two GORM models
- add migration 000007 dropping ccg_postprocess_policy_state and
  ccg_postprocess_run_logs (sqlite + postgres); RequiredSchemaVersion 6 -> 7
- remove get_postprocess_policy and reset_postprocess_policy MCP tools (26 -> 24)
- drop the postprocess_policy request arg and fail_closed path; keep the
  inline degraded behavior (warn + failed_steps) in build_or_update_graph and
  run_postprocess
- remove the PostprocessPolicy Deps field, runtime wiring, HTTP /status
  postprocess summary, and the `ccg status` postprocess section (+ --errors/--recent flags)
- delete guide/postprocess-failure-policy.md (EN + ko) and sweep doc references

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@tae2089
tae2089 merged commit 7fd62e3 into refactor/remove-quality-analyzers Jul 11, 2026
1 check failed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant